A method for selective SVM integration based on cultural algorithm and negative correlation learning

In this paper, a method for selective SVM integration is introduced in order to improve the generalization performance of SVM, which is based on cultural algorithm and negative correlation learning. This method mainly includes four parts: independent sub-SVMs training by bootstrap technology, creating an adaptation function based on negative correlation learning, computing the optimal weight of SVM in the weighted average values, and SVM integration with the weighted value which is more than a given threshold value. In the experiments, this is an efficient and effective method to improve the generalization performance of SVM.